In this episode, we chat with Industry Veteran Karl Isler of Karl Isler Consulting. Dr. Isler talks with us about dynamic pricing and his recently published whitepaper, “The Art of the Possible in Dynamic Pricing.” Dr. Isler also describes his passion for solving what he calls “interesting problems” in revenue management.
Dr. Isler feels that dynamic pricing is a superior type of control that is often misunderstood in the industry. As a freelance consultant, he sees mishandling of sell-up fairs and believes that if airlines learn how to handle sell-up fairs properly, it helps them master dynamic pricing.
In This Episode
[0:45]: Karl Isler’s Journey in Aviation: From Physics to RM
[5:00]: Major Milestones in RM: From Leg to O&D
[6:40]: What really is dynamic pricing?
[10:00]: The Art of the Possible with Dynamic Pricing
[13:12]: What are airlines trying to solve with Dynamic Pricing?
[18:10]: What can success look like?
[20:05]: Taking it to the next level: continuous pricing
Aditi: Hello and welcome to the PROS travel podcast series, The View from 30000 feet. I'm your host Aditi Mehta. In this episode, Justin Jander talks to Karl Isler from Karl Isler Consulting about dynamic pricing and the art of the possible. Let's take a listen....
Justin: Hello, and thank you for joining us today. I'm excited to host today's podcast. My name is Justin Jander. I'm a product manager here at PROS in our revenue management products. I'm excited to be joined today by Dr. Karl Isler and he's going to tell us a little bit about a white paper that he's recently published, as well as some of his findings throughout the journey of revenue management. So, hello Karl.
Karl: Hello, Justin.
Justin: All right, so I guess to get us started, can you tell us a little bit about your journey in the aviation industry? I was just telling somebody the other day that Karl, you're an expert and a legend in the industry. You've been around for a long time and seen the industry grow and change. So if you could tell us a little bit about where you started in the industry and how you've gotten to where you are today?
Karl: Well actually I started out in a different industry. I used to work in research in elementary particle physics, like field theory and stuff like that ...
Justin: A little different from I think revenue management.
Karl: It's a little different but then sort of I found that they were not enough interesting problems to work on. I mean there are grand problems but the chance that you can solve it is rather low. So there were no real problems around. So I eventually I quit that field and then I pretty soon afterwards I saw a job announcement that Swiss Air where they were looking for somebody in the field called revenue management. I didn't have an idea what this really was. But when I came there, it was really fascinating because a lot of elements of particle physics, like particle creation and particle go aways, like bookings appear and go away. So actually I find myself quite at home.
Karl: And also how the things work. It was kind of like an inner CVS research. They've had conferences you could publish papers and so on. And the nice thing about it was also that there were problems just jumping at you from all the corners. And so basically I got stuck in revenue management and then we had a unique chance also to influence quite a lot how things go because after Swiss went bankrupt, Swiss started out as more like a startup company feeling and we could together with PROS actually build the foundations of many things which are still around, call it dynamic pricing or [inaudible 00:02:35], all these modules. That was actually the start of quite a long relationship we had with PROS so we could actually do some innovations. And that was really exciting.
Justin: I find it interesting because we have several people at PROS that are former physicists and even a few nuclear and particle physicists that sort of all have gravitated to this field. And it's interesting to see how you can take a very science field and apply that science to a business setting. And as you said, the ... I think one of the things that's interesting is that the problems constantly changing as the world changes and it's whereas physics, I think you can probably research one little thing forever.
Karl: Yeah. I mean usually you have what do they call a standard model in physics, which is pretty good and it's pretty tough to even bring something better there.
Justin: Well as we certainly know from PROS that things do change a lot. And of course, we have evolved from our Leg based system to our O&D based system. And you talked about when Swiss air came around, Swiss air was the first O&D customer, PROS O&D customer and I believe you had a pretty big hand in the development of that. Can you tell us a little bit about that?
Karl: Yes. I mean that was actually, it was a project which survived the company. The project started out with Swiss Air, which was grounded in 2001 and then for a while it was there, but then the new company was formed, which is called Swiss. Just lost the air. But then we could actually bring the project over into Swiss. And the nice thing about it was that a lot of things had already been done in the Swiss Air area so we could actually sell it to the new Swiss company, which was supposed to be a much smaller company, but we still could implement the OD system. So we were actually quite actively working with PROS to make this thing work and that'd we had a relation that we at the Swiss side would also look into the system and test the data. We would propose solutions. And one thing which I really never forget is that ... it was about four months before cut dollar, we came up with an idea to modify things in RTDP, the calming is this fair chastisement things. And then luckily this was possible four months before cut though of such an important project that this really somewhat exceptional for industries. I mean in more bigger companies, I think you couldn't do that.
Justin: So just to help us kind of proceed through where we are today. And again, I mentioned your white paper and we'll get there. When you talk about going from Leg to O&D, what was the real kind of motivator for that? I know a lot of people don't hear O&D today and it sort of accepted. What was sort of the reason behind that at the time?
Karl: I would say, I mean if you look at the network of Swiss air, I mean Switzerland lies quite in the middle of Europe and it's a quite a good hop position. You have also another thing is we have in Switzerland is you have inbound and outbound traffic, I mean there's quite some industry in Switzerland, so we have a lot of business traffic inbound. Also, the industry situation, business in Switzerland is actually good for having, they want a lot of long haul flights to be better connected, but you can support the long haul flights only by doing a Europe connection.
Karl: So the network structure was quite optimal and so that was the original motivation that we say well we should actually do network control instead of flight control. We can make, I don't know what 1.5% more revenue. So we did that with simulations. Where is a better barber and MIT and stuff like that. In the end it turned actually out that and not that aspect was even more important with OD, and this is that you need what they call seamless availability. That means you have a basically possibility to inference the availability on the fly.
Karl: And that turned out to be a much better revenue driver because it enabled us to do what we call dynamic pricing, which is might be a little bit different from what the industry today, the airline industry today thinks what dynamic pricing really is.
Justin: Yeah, and I think that really gets into the next part. So, PROS has had ... you've referenced PROS RTDP a few times and that stands for real time dynamic pricing. And so of course the industry buzzword these days is dynamic pricing. When are we going to have it and how is it going ... what's it going to look like? All of those things. But of course we've had a product called real time dynamic pricing since I guess 2005, 2007 something around there. Right?
Karl: Yeah, the precursor was already in the original, 2003 module actually had.
Justin: And so we've continued to evolve that into what we have today. So can you tell us a little bit about ... you were a big part of it becoming real time dynamic pricing and obviously there's the component that's the real time and there's the component that's the dynamic pricing. So can you elaborate a little bit on what dynamic pricing meant back then so that we can sort of help establish the baseline of real time dynamic pricing in 2003 to today versus where it's headed in the future?
Karl: Well, I mean what I personally believe is that it's a little bit misunderstood in the airline industry nowadays, what dynamic pricing should be, they confuse it a little bit with let's say what we call continuous pricing that you have to file fares first and so on and so on. Economically actually dynamic pricing is just that you can change the price for something on the fly, and of course prices change. I mean the end of the NSC you have price changes but many of the price changes are because different. You ask different things. If they ask for a round trip, no minimum stay fair. Just a different price than when they ask for something.
Karl: So some of the prices which is not 3d dynamic comes from the selection that you select for different customer, different prices. The true dynamics of the prices actually comes from what in the older world is called the availability in the newer world is maybe called the bid price is that your estimate, how much resources you have kind of changes. And is also very stochastic, if you have a large group booking, which you didn't know before, suddenly you have much less space than you thought you had. You have to put your prices up and down and all this. And I think that the real thing about this, I would say about dynamic pricing is that you can change prices for the same thing based on your resource situation.
Karl: And actually airlines did that maybe not knowing really what they're doing so well ... did this all along because I mean how would you as an airline change the price for the same thing? And essentially because you have to file all the fairs with ATP code.
Karl: So that's not so easy. You cannot really refile it for each occasion. But what Danielle has came up with they say, well, let's just for the same thing file a little bit higher fare, but a bunch of them and then just availability. We can switch them on and off. And this is really a way to do the [inaudible], pricing because you can change the price for different things. You can only do it in discreet steps, by the way. But I think that's not so much of a disadvantage. Of course, it would be better not to having to file all these higher fares, but essentially if you file a really decent set of higher fare levels and are able to control them ...
Karl: Then you can do dynamic pricing and that is exactly what we did with RTDP. We basically tried to control the availability of those so-called by up levels.
Justin: Right. Yeah. It's such an interesting thing because as dynamic pricing, the term has become more and more popular. People are starting to use it interchangeably and sort of have left the RTDP type of dynamic pricing has sort of just been pushed to the back burner and everybody's talking about continuous pricing and so forth when there's still a lot of improvements that can be made to how you do the dynamic pricing that's available in RTDP. So yeah. As we sort of move from the kind of talking about how dynamic pricing the term has evolved. You wrote a white paper recently, it was called the art of possible and dynamic pricing. It's an interesting title. Can you describe a little bit about what you meant by the title and why you wrote it?
Karl: Yes. I think that came a little bit from my experience I have now as a freelance consultant. I basically quit working for one company only and do some freelance consulting. And I see now how other airlines actually deal with the problem of setup fairs and so on. And I note that there is a lot of misunderstanding what these things should be and how to deal with it. So as I said before, I think that even being able to handle setup levels properly is a form of dynamic pricing. And this is why I call it the art of the possible, because I think airlines would first have to learn, to deal is to set up first properly. Once you can learn to set up first properly, it's like you solved the dynamic pricing problem with the restriction that you price can only be one place out of a bunch of discreet prices.
Karl: It's an additional restriction. Now you can say is this restriction really a bad restriction or not. Well, it is a restriction, but I claim most of the benefits you already get. If you can handle four or five setup levels, and this is what I mean by the art of the possible. That means if you really learn how to handle these four or five setup levels properly, it's a form of dynamic pricing which you can do now. It is not really restricted by having to do change the whole distribution mechanism and so on. Of course, it's still restricted in one or the other ways, but I believe airlines should first learn how to deal with that.
Justin: So it's sort of as you learn how to deal with the first problem, it helps you then enter the second problem rather than just skipping over the first.
Karl: Yes. Essentially you can also then interpolate that ... let's say you mastered the discreet problem is the discreet set up levels you mustered a properly, that gives you immediately a mathematical handle, how to do the continuous prices. This is actually what we did at Swiss with the groups because groups was not tampered by the distribution issue. So we say why not channelize this discreet one into the continuous problem. And we were probably the first airline to come out with true dynamic pricing underneath the carrier side.
Justin: And just as a note that was with PROS, GSO, that you did that part. So important to note there that implementation was with our group sales optimizer tool, which we're now taking and extending beyond just in groups and getting into the individuals.
Karl: Yes. The idea there was, because groups are not really filed group president are not being filed, they were quite basically caught at a talk.
Karl: So that was possible.
Justin: So one of the things you said when you were talking about your paper is that you're starting to visit a lot of different airlines and you're getting that experience around the world of how this is approached. There's a lot of philosophies on dynamic pricing, but if you can explain sort of what is the crux of the problem that the airlines are trying to solve with dynamic pricing that I think will be helpful?
Karl: Well, on one hand, I think dynamic pricing is just a better, a superior type of control because in the old 11 dimensional model, you would just switch on and off demand streams, right? If you see you don't have enough capacity, you just say this demand stream should be closed, but this is not really very optimal control because if you think one demand stream has different customers with a little bit different willingness to pay. Of course, they're all like segment and have maybe similar willingness to pay but still they have different willingness to pay. Instead of closing off the demands, the mentality, why don't you just raise the price a little bit. So there's still a ... some customers we've seen that demand same meeting, they might buy it.
Justin: Yeah. I think that's a really interesting point because I know from my experience of visiting airlines, that is a very easy way to do it. It's sort of intuitive to just say, I want to force someone to buy a higher class just close off the thing that they would have bought. But, to your point, there's a much more mathematical approach to saying change what the value they provide to me is so that then they're worth more to me and I not just a blunt tool, I'm giving more of a mathematical tool to it.
Karl: Yes. I mean the old model of independent demand would only work if each customer has the same willingness to pay, but this is not real. In reality, you have to assume a certain distribution of willingness to pay. So that means even if you're a certain passenger segment, you is still take the upper part, you just that take only those with a higher willingness to pay on the given segments rather than closing down the entire demands stream at once so is a real better way to control. I mean I think the islands realized that pretty soon because of there ... that is probably also one of the reasons why they were never really happy with the availability in the old revenue management system because if you just think in standard availability, it cannot bring in all those pressing dimensions which you might have. I mean that you say maybe short before departure minister pays higher, therefore they close a lower class, they do set up all these things do not come naturally if you're not imposed sort of the dynamic pricing aspect to it.
Justin: Right, and that's where you can say that there's sort of two components to how you do revenue management. You need a good bid price that sort of represents the volume of the demand that's out there and the value of that demand. But then you also need the secondary component of, now that you have the bid price, that sort of tells you the capacity level. Now what about the willingness to pay around that bid price?
Karl: Yes, that's a very important point. Doesn't mean ... economically in the backend there's really two mechanism what can happen. And one mechanism is, given that customer segment, what is his willingness to pay what you want to charge to him? Right? That's one aspect is the pricing aspect. But the other aspect, even if you would charge the optimal price to each customer, you might not have enough capacity to really take all those passengers that is the bid price aspect which says while I cannot transport everybody so I have to impose like a congestion surcharge to the price so to speak, to slow down the demand on that given flight, which has a constraint so that it fits capacity.
Justin: I think that's a really interesting way of describing what a bid price is, a congestion surcharge. I think that's pretty interesting. I might have to steal that for future use, but it is and that's something we do often sort of run into is explaining what a bid price really is because a lot of times people in kind of combined the two a bid price combined with the willingness to pay part and that's really something that they need to be separate. They're separate economic pieces and they should be modeled so.
Karl: It is very important. Yes, and that's one of the ... I think one of the problems most airlines have is this fighting by dancing. Let's say if you're on a locus cat and you just have one leg or demand local demand on one leg and zone, it doesn't really matter whether you add them together, you put this, the willingness to pay piece together is the bid price and so on, but in an all the world it's extremely important that you separate these two things. I would call you need two controls. In an OD world, you need the bid price is like the congestion control, but you need the pricing control, which is the margin you want to put on top of the bid price. Now the problem is if airlines tried to forget about you having to put the margin, they might inflate the bid price artificially and so on, but the problem is then if you try to add bid prices on two different legs in a connection, your bid price is too high. The margin is not additive. Whereas the bid price is additive and you have to keep those things apart. What is additive and what is not. If you start to add two margins together, you get what we call double marginalization and that is never good.
Justin: All right, so we've talked a lot about the kind of math behind it, the economics that drive it. I think one of the things that's important too is how users interact. So from your experience with airlines around the world, how have you seen success in how the users take those two economics approaches and make it successful at the airline?
Karl: Well, I would have to say what we tried at Swiss, I still find one of the most successful ways of doing these things. These has been extended to the entire Lufthansa group by now and what we came up with essentially say there are two economical principles going on. One is congestion control and the other is kind of a pricing margin. You should assign two separate roles to those two things. Essentially we have a price so you could maybe it's very similar to the traditional price except that the price, it would not determine the absolute fare level. It would more determine how much you should add on top of the bid price, and that's the price role. And then you have the capacity control role or whatever you call him and he is responsible for the bid price. The bid price has to do more like with future demand, future bookings, it's about the future.
Karl: How much can you make with that seat in the future? Whereas the pricing control has to do ... how much is the current customer in front of you willing to pay? And these are by nature, two different things. So they also assigned two different roles to that and this has to turn out to be quite successful. And one of them ... all the important points is if you are in a network, usually the bid price comes out of a network optimization and so on. And the problem usually then is nobody's really responsible so much who is responsible now for the bid price. What you also find very useful is to really make the leg capacity controller be responsible for the bid price so that he would be able to change that bid price in case it is necessary. And he's fully responsible for that.
Justin: Makes sense. So we've sort of covered a variety of different topics here and focusing a lot on the art of what's possible in dynamic pricing. And now you're, as we think about it and we move towards in the system of systems, evolve into what people are saying dynamic pricing is today. Can you talk a little bit about the science and the users as we really do move to that next level of continuous pricing?
Karl: Well, I wonder if the proposal was that those two different roles would also have to deal with two different, let's call forecasts. Maybe they're not both like forecast, it just tried to put light on different aspects of demand, right? The pricing forecast on we call it more like an estimation is rather a scientific tool trying to numerically estimate the willingness to pay so to speak or some call it price elasticity. In principle, what you need is the probability that you get the sell at a certain price. That is basically what the prices should know. On the other hand, you have to capacity control it. This is more about the future DCS all what the future uncertainties, what can going to happen, so this one is the resource role and the resource role should actually predict how much money you can make with deceit otherwise and sort of the online calculation and tries to gauge those two things.
Karl: It says how high should I charge to the current customer with the risk to lose the sale, compared to what I would make otherwise. Right? This is kind of the balance which has to be calculated. And that needs to kind of two scientific kind of models. One is more about future forecasting stuff, time series model and so on to come up with a decent bid price. And the other model is more like, can I estimate based on my customer information, whatever I know about the customer to what kind of segment he would belong and what is the best possible price for that customers. So they are they kind of two disciplines. And to put that forward, I believe we need to change the way how we think about the pricing. I mean you cannot at the same time file very complicated, [inaudible 00:22:11], and on the other hand try to have a scientific understanding about willingness to pay. That's somewhat exclusive. So dynamic pricing on the bottom line would actually be very simple because what would you find a price be? You would just add up all the bid prices of the resources, the customer wants, you add the pricing control on top. There you go.
Justin: Yeah. Well that's interesting and it certainly sounds simple and now it's just a matter of getting all of the extra stuff that goes with it together and all the different components of the systems that have to support it, which is again why it's good that we have the approach of dynamic pricing that you talked about at the beginning to where we're headed.
Karl: Yes, because that's very important because one of the big roadblocks is actually the current distribution model, specifically indirect distribution. You say now a lot of airlines, now they do let's say up to 50% direct distribution, but the airlines didn't really inventing a completely different pricing mechanism. Still, if they do direct distribution, they still apply the old indirect distribution model to calculate the prices. And that has been one of the biggest roadblocks because also the industry is interlinked. There's a so called network effect. As Swiss maybe we could do dynamic pricing, but we have a culture. We have a joint venture and then we would have to partner with the other airlines and then the other airlines would not be able to do that, so it's all interlinked and to break that vicious cycle, it's very difficult and that's what we tried actually is the so-called NDC new distribution capability should be a way to break that, but the introduction and adoption of new distribution capabilities specifically on the indirect channel, is rather slow.
Justin: Yeah, and I think that's why it's important that just because those roadblocks or hurdles come up, doesn't mean that there's not options before those get there.
Karl: That's the point of the white paper, the art of the possible. I say you can do dynamic pricing almost in the current environment. You should do it in the current environment because this is how you learn how to do dynamic pricing and then you are also fit to go to the new world of the continuous prices.
Justin: Perfect. Well, Karl thank you very much for your time today. It's been really enlightening and really interesting to hear from you, and I know the audience is really going to be interested in it and interested in reading the paper as well. So thank you very much.
Karl: You're welcome.
Aditi: Thank you for listening. And a special thanks to Karl Isler from Karl Isler consulting for taking the time to talk with us. This podcast is brought to you by PROS travel. At PROS, we help airlines on their journey toward offer optimization and digital transformation. For more information, please contact PROS.